http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44107
Iain Sandoe <iains at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target|powerpc-apple-darwin9.8.0 |*-apple-darwin{8,9,10}
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2011.06.26 17:37:00
Component|bootstrap |target
CC| |dominiq at lps dot ens.fr
Host|powerpc-apple-darwin9.8.0 |*-apple-darwin{8,9,10}
AssignedTo|unassigned at gcc dot |iains at gcc dot gnu.org
|gnu.org |
Ever Confirmed|0 |1
Summary|libstdc++ (dylib) is built |gcc emits frame (epilogue)
|with an erroneous |info incompatible with the
|dependency towards /usr/lib |darwin
| |{8,9}-unwinder,10-compacter
Build|powerpc-apple-darwin9.8.0 |*-apple-darwin{8,9,10}
--- Comment #16 from Iain Sandoe <iains at gcc dot gnu.org> 2011-06-26 17:37:00
UTC ---
This bug has been around since part way through 4.5.
Unfortunately, it is somewhat hidden in the test-suite since the latter
interposes the newly-built libgcc ahead of the system version (because
DYLD_LIBRARY_PATH is set to point to ./gcc, of necessity).
There are two steps needed to deal with this fully:
(a) fix things so we don't emit epilogue info in unwind frames for darwin 8,9
and 10 (compacter).
The patch currently attached is OK on 4.6.x - but the emitting of unwind frames
has been recently vec-ified which means it will need some re-work for trunk.
(b) fix the darwin libgcc build process so that we only export the _ext symbols
from newly-built libgcc when this is only supporting the ext.
* (b) will also solve the library ordering issues when -{force}_flat_namespace
is used.
I have another patch in progress that is a first attempt at (b) - (not yet
ready to post).
